Kumaresh Bhuyan · Head of Programme & Operations

Programme leadership · Kreeda Labs

Kumaresh Bhuyan · Pune, India

I'm Kumaresh Bhuyan. Ten years in technology delivery, from the QA floor to the programme office. Today I run enterprise AI, GenAI and SaaS portfolios at Kreeda Labs — shipping for global brands across these seats:

Luxury fashion
BFSI
Retail
FMCG
EdTech
Public sector

Work Journey All cases

Field note: the frontier gap is a definition, not a finding

The Execution Edge · 601 subssubscribe →

Now: Programme leadership · Kreeda Labs, Pune
Focus: Enterprise AI · GenAI · SaaS
Proof: 30+ systems · 2,600+ followers · PMP

Work Journey All cases

years from branch floor to programme office

portfolio efficiency after the PMO I introduced

0M+

players on live titles I helped ship

project completion lift under those frameworks

About

Scaling Enterprise Technology & AI Products · Engineering Delivery · Strategy to Execution

Strategy becomes outcomes in execution.

I'm Kumaresh Bhuyan. Ten years in technology delivery, from the QA floor to the programme office. Today I run enterprise AI, GenAI and SaaS portfolios at Kreeda Labs, shipping for global brands across luxury fashion, BFSI, retail, FMCG, EdTech and the public sector.

Execution is where strategy becomes business outcomes. Helping organizations bridge that gap has been the focus of my career.

Over the past decade my career evolved across quality engineering, product delivery, programme management, and operations — an end-to-end view of how technology products are built, delivered, and scaled.

Today at Kreeda Labs I lead enterprise technology and AI initiatives with engineering teams, business leaders, and clients — clarity, accountability, and measurable outcomes.

Sustainable execution is built on strong teams, clear governance, practical decisions, and continuous improvement. I build delivery environments where people perform and organizations scale with confidence.

My work has centered around

Leading enterprise technology, AI, GenAI, and SaaS initiatives from strategy through delivery
Strengthening PMO governance, delivery frameworks, risk management, and executive reporting
Building engineering capability through hiring, mentoring, and workforce planning
Aligning business priorities with technology execution across cross-functional teams
Delivering across BFSI, Government, Retail, FMCG, EdTech, Gaming, and Enterprise Technology

Organizations scale when execution becomes a shared capability — strong leadership, disciplined processes, and empowered teams.

All Star Games

Led live mobile-game delivery, including a title that hit #1 Top Free and reached 3M+ players.

Kreeda Labs

Lead enterprise AI and technology programmes for enterprise and public-sector clients.

Core areas of expertise

Enterprise Technology DeliveryAI & GenAI ProgramsEngineering DeliveryPMO & Delivery GovernancePortfolio & Program ManagementStrategy to ExecutionEngineering OperationsAgile DeliveryOrganizational ScalingExecutive Stakeholder Management

Current seat

Programme leadership · Kreeda Labs

Focus

Enterprise AI · GenAI · SaaS programme delivery

Path

Banking ops → QA → PM → Programme leadership

Voice

The Execution Edge · field notes & essays

If you are building scalable technology organizations or solving complex delivery challenges — let's talk.

Contact LinkedIn

Full career timeline

Programmes by sector

Real projects. My delivery seat on each.

Drawn from programmes I ran or owned end-to-end — NDA keeps brand names off the page; outcomes and stakes stay.

Full archive

Fashion AIOpen →

Material visual search

100K+ assets · habit change > model score

3 hrs → 3 sec

HR TechOpen →

8-agent HR assistant

Policy answers under legal + IT sign-off

70% fewer queries

HealthcareOpen →

Spinal segmentation

Scarce labels · clinical trust bar

90%+ Dice

GovernmentOpen →

Public stats chatbot

Four census cycles · audit trail first

Hours → seconds

Selected work

One operating seat. Five projects — by magnitude of ownership.

Flagship is the PMO seat I run today. The grid packs around the Deftouch / RCB card — no empty rails under shorter cases. Client brands under NDA stay unnamed.

Full archive →

01Operating seat

Flagship · Current seat

The programme office that made AI programmes survive handoff

Open case

What I owned. I built and run the PMO: governance, capacity, risk, reporting, and the rituals that force a named owner before anything ships. Strategy → executable programme → production with receipts.

What moved. Portfolio efficiency up 23%. Project completion up 32%. The same operating system now sits under every engagement below.

23%

efficiency I unlocked via PMO frameworks

32%

completion lift under that system

Multi-team

programmes I oversee end-to-end

Hiring

end-to-end technical recruitment I own

02MajorNDA · sector only

Major · Computer vision programme

Material discovery where adoption beat the model

What I owned. I owned the delivery runway: scope gates, stakeholder sequencing, and the definition of done that forced the tool to beat old habits — not just beat a benchmark.

What moved. Discovery collapsed from hours to seconds at 95% detection accuracy across 100K+ indexed materials — with a path to trust the team would actually use.

3 hrs → 3 sec

discovery time we had to beat

95%

detection accuracy at handoff

100K+

materials under the index

Open case

03MajorNDA · sector only

Major · Agentic delivery

Eight HR agents — and a six-week trust runway

What I owned. I ran the cross-functional runway to production: agent boundaries, review gates, and the six-week clock that forced decisions instead of endless pilots.

What moved. 70% fewer repetitive HR queries. Eight specialist agents live. Six weeks from design call to production under real policy risk.

70%

fewer repetitive HR queries

specialist agents in production

6 wks

runway I held to go-live

Open case

04Foundation

Foundation · Deftouch / All Star Games

Live-ops at Deftouch — chart pressure, ratings, and a studio that scaled

What I owned. At Deftouch (All Star Games) I grew from Lead QA to Associate PM: test plans → pipelines → end-to-end schedule, hiring, and risk for the cricket run. The scaling work was concrete — helping take the studio from five people to forty-two while keeping release discipline intact under Nazara / RCB chart pressure.

What moved. RCB Star Cricket held a 4.4★ store rating under live fire, hit #1 Top Free in 48 hours, and helped the slate reach 3.2M players. Nine titles in the run; five reached real players. That Deftouch operating habit is still how I run AI programmes.

Titles from the Deftouch run

RCB Star CricketCricket StarAll Star CricketAll Star Cricket 2Zen TripleTile Match 3DBaseball Smash

4.4★

RCB store rating under live pressure

Top Free chart within 48 hours

3.2M

players on titles I helped ship

5 → 42

Deftouch studio scale across my years

Open case

05FoundationNDA · sector only

Foundation · Public sector programme

Public data chatbot with an audit trail first

What I owned. I insisted the audit and explainability path ship before conversational polish — four census cycles of source data, zero appetite for untraceable answers.

What moved. Query response moved from hours to seconds against a gov-grade traceability bar — with the audit story intact at handoff.

Hours → sec

query response we had to beat

Gov-grade

traceability bar I held

4 cycles

census source depth in scope

Open case

06FoundationNDA · sector only

Foundation · Healthcare AI programme

Spinal cord segmentation under scarce labels

What I owned. I kept clinical trust criteria visible in every gate while transfer learning closed the accuracy gap — validation discipline equal to the model work.

What moved. 90%+ Dice (~0.9) under scarce labels, with a healthcare buyer who could sign production because the validation story was as strong as the network.

90%+

Dice ~0.9 accuracy at handoff

Transfer

learning under scarce labels

Clinical

trust criteria I held as gates

Open case

Also on the programme

Full archive with detail →

Logistics

Trucking marketplace delivery

Customers, transporters, drivers on one platform — I kept ops milestones honest across three personas.

Maritime

Ship & port risk modules

Real-time captain insights — programme glue between analytics and bridge workflow.

EdTech

Script generation platform

60–70% faster production once content, research, and rewrite gates stopped thrashing.

Fitness · Gaming

Connected mat + guided workouts

Live product ops lessons that still shape how I staff AI programmes.

Creator Economy

Anti-piracy monetization

Creator income 3× in three months — delivery under messy multi-party incentives.

Confidentiality, by design

This site is my operator record — not a mirror of any company marketing page. Sectors, problems, and measured outcomes appear; client names under NDA do not.

Operating line · From my seat

Deployment is a purchase. Absorption is a project — and I own the project.

The line behind every programme I run.

Career path

Seven seats. Ten years. Ownership earned floor by floor.

Early seats stay compact. Recent seats open as chapters — what I owned, what moved.

Apr 2016 – Jun 2017

Branch Development Manager

State Bank of India · Bhubaneswar

First leadership seat without a slide deck to hide behind.

Aug 2017 – Mar 2018

Freelance QA

Hyderabad

Quality is a process, never an event — still the standard I enforce.

May 2018 – Mar 2019

Lead QA / Project Associate

Deftouch Interactive Art · Bengaluru

Built the engineering trust that made delivery ownership possible.

Jul 2019 – Mar 2021

Jr. Project Manager

Deftouch Interactive Art · Bengaluru

Owned the gap between intent and sprint.

Jul 2024 – PresentNOW

07 / 07

Head of Programme & Operations

Kreeda Labs · Pune · On-site

Lead programme and operations for enterprise technology and AI portfolios — partnering with engineering, business leaders, and clients on complex initiatives.

Portfolio reporting, financial tracking, capacity planning, and operational reviews
Engineering capability via hiring, mentoring, workforce planning, and performance development
Executive stakeholder communication, client governance, and strategic escalations
Continuous improvement of delivery practices as the organization scales

Why this seat mattered

PMO governance and portfolio systems that lifted execution efficiency by 23%.

Kreeda Labs · Pune · On-site

Lead programme and operations for enterprise technology and AI portfolios — partnering with engineering, business leaders, and clients on complex initiatives.

Portfolio reporting, financial tracking, capacity planning, and operational reviews
Engineering capability via hiring, mentoring, workforce planning, and performance development
Executive stakeholder communication, client governance, and strategic escalations
Continuous improvement of delivery practices as the organization scales

PMO governance and portfolio systems that lifted execution efficiency by 23%.

Credentials

Four highlights — full list with details and learnings on the certifications page.

All certifications

Certification

Jul 2023

Project Management Professional (PMP)

Project Management Institute

Professional Certificate

Jan 2025

Microsoft Azure AI Essentials Professional Certificate

Microsoft and LinkedIn

Certification

Aug 2023

Career Essentials in Generative AI

Microsoft and LinkedIn

Course

Jul 2025

AWS Certified Cloud Practitioner (CLF-C02) Cert Prep

LinkedIn Learning

Writing

Long essays and near-daily field notes, from the delivery seat.

Deep dives on what enterprise AI adoption actually takes — plus short operator takes on the week's news.

Long-form · Essays & The Execution Edge newsletter

Operating Rule · New

The Symmetric Value-Gate Rule.

Your process asks one question, and it is not what a thing costs or commits you to. It asks whether the thing presents the trigger, and almost nothing consequential has to answer that. Audit what makes your gates fire, not what they were written to cover.

Read the essay →Approval Gates · Governance

Field Essay

The software you just decided to own.

Nearly a third of organisations have stopped buying software they would previously have bought, and built it instead. The build clears every threshold the purchase would have tripped, so a permanent maintenance obligation arrives with no review behind it.

Read the essay →

Field Essay

The frontier gap is a definition, not a finding.

OpenAI ranked enterprise customers by tokens per active user, called the top ten percent frontier firms, then reported that frontier firms use more tokens. The skills and plugins numbers in the same report are the ones that hold, because their denominator is fixed.

Read the essay →

Field Essay · On this site

Cancelled AI projects signal discipline.

Three enterprise AI numbers read as evidence the technology is stalling. Read again, they show buyers finally pricing AI on proof of value. The gap worth worrying about sits in the platform bets nobody gates.

Read the essay →

All essays, published here first →

The Execution Edge · 601 subscribers and growing

A monthly deep dive on enterprise AI and turning strategy into execution. Edition 01, The Execution Illusion, is also a designed PDF: read it offline or drop it straight into your next steering-committee deck.

Subscribe on LinkedIn Download Edition 01 (PDF)Substack Medium

Field notes · Posted almost daily

Today's Note

The next note is probably already up.

A new field note lands most weekdays. Following on LinkedIn is the fastest way to catch them the day they publish.

Follow on LinkedIn →

Remediation SLAs

Your severity bands outlived the benchmark under them.

CISA retired the model on 10 June. BOD 26-04 drops severity rankings entirely and keys a federal deadline to public exposure, presence on the KEV catalog, how automatable the exploit is, and how much control it hands over. That reaches delivery before it reaches security.

Read on LinkedIn

Absorption Layers

The constraint is almost never the model.

Four layers, checked in order: deployment capacity, cost and compliance defensibility, portability, and organisational absorption. Meta's Project OT is the expensive worked example, with code changes up 220 percent against features reaching users up 36 percent.

Read on LinkedIn

Open Source Commons

Nvidia did not buy a model repository.

A 12.9 billion dollar acquisition of Hugging Face at 86 times revenue is not a financial calculation. The neutral standard-bearer for the open developer community now sits under the dominant compute vendor, which rewrites the social contract rather than the software moat.

Read on LinkedIn

Agent Reversibility

Sort agents by what you cannot undo.

Autonomy levels tell you how often a human checks, and nothing about what happens when that human is wrong. The Reversibility Rule asks what an agent can do that you could not undo within an hour, with one person, without calling the vendor.

Read on LinkedIn

Signal Dilution

Ten million credentials, thinner signal.

India's Red Fort target trains ten million young people in AI within a year. That raises the supply of credentials, not the supply of people you can trust with a production system — and your screening question quietly stops separating anyone.

Read on LinkedIn

Want something built for a steering committee, not a scroll? Grab the 1-page AI Governance Checklist ↓.

Free resource

The 1-page AI Governance Checklist.

Global frameworks set a floor years away. These five controls protect your AI programmes this quarter.

Audit rights before signature

Evaluation rights in every vendor contract before signing — inspect the model, do not merely trust it.

Kill criteria, set in advance

Define the failure that pulls a model from production while everyone is calm.

One named human per decision

A single accountable owner for each automated decision. Accountability cannot sit with a system.

Log everything, and the why

What the model did and why — explainable later to a regulator or a customer.

Review the boring failures first

Quiet failures hide silent drift. Review them before the loud incidents.

Recognition

Where the work gets noticed.

Field notes, practitioner threads, and the lines people quote back — not a trophy wall.

Featured in

Humans of GamingGaming · Deftouch

Profiled in the "Humans of Gaming" series on the run of real-time multiplayer cricket titles shipped at All Star Games (Deftouch).

Read the feature on LinkedIn →

Kreeda LabsEnterprise AI · Now

My current seat: programme leadership for enterprise AI and SaaS delivery systems. Listed on the company leadership page for verification, not as a brochure mirror.

Verify role →

Lines people quote back

Agent Write Access
“Never let a plausibility engine hold the pen on anything irreversible.”
Acknowledged in-thread by the post's author

Escalation Ownership
“Observability without a named escalation owner just produces a detailed record of a mistake nobody stopped in time.”
Acknowledged in-thread by the post's author

CX Ownership vs. Incentives
“The harder problem isn't who owns it, but what they're measured on once they do.”
Acknowledged in-thread by the post's author

Technical Debt
“Most debt is created at kickoff, when the deadline locks before the dependencies are mapped.”
Acknowledged in-thread by the post's author

Operating Model Ownership
“The operating model change doesn't get skipped because nobody understands it — it stays untouched because it costs someone their authority.”
Acknowledged in-thread by the post's author

AI Sign-Off Ritual
“The part that breaks first in my experience isn't the testing process, it's the sign-off ritual built around it.”
Acknowledged in-thread by the post's author

Threads that weren't mine to begin with

Redesign vs. Tooling Authority
“The gap between that question and what gets funded is where this usually breaks. A tool purchase has one owner who can sign it off. Redesigning how the enterprise operates cuts across P&Ls that no single leader controls, so the ambitious version gets nodded at in the room, then quietly re-narrowed into a tooling decision, because that's what someone can actually approve.”

SOP Maintenance Ownership
“The automation and agent split is right, but there's a maintenance problem underneath it that rarely gets budget. Someone has to keep noticing when the SOP goes stale as the business changes, and in most programmes that job doesn't exist — it just quietly becomes nobody's. An agent scoring itself against an outdated rulebook still looks confident.”

Agent Identity & Audit Trails
“Most enterprise agents inherit whatever service account or API key the engineer who built them happened to use. So when one goes off and issues a 20 percent discount on its own, the audit trail points to a person who never approved that spend. The fix isn't better fallback logic, it's giving the agent its own identity from day one.”

Framework Choice vs. Ownership
“Most enterprises pick LangGraph or CrewAI before anyone has agreed who actually owns the audit trail those frameworks generate. Six months later the tool gets blamed for a governance gap that was really an org design call nobody made.”

GCC vs. ODC Governance
“The gap isn't that a GCC needs more governance than an ODC did, it's that the old model served one relationship, the parent company deciding what to delegate. A GCC has to answer to several owners at once — strategy, IP, product, and risk — each with a different definition of what "working" means. That's the redesign most orgs skip, because the review cadence never fails loudly, it just quietly stops representing anyone.”

Contact

Tell me what you're trying to ship.

Programme leadership, enterprise AI absorption, or a straight conversation about what adoption takes. Inbox only — no phone gate.

If you're building scalable technology organizations or solving complex delivery challenges — I'd be glad to connect.

kumaresh@kreedalabs.com kumareshbhuyan301@gmail.com

Or follow the field notes

Follow · 2,600+Subscribe · The Execution Edge